What a proof is for, and what it is not for
A proof is for errors introduced since acceptance, and for errors of fact that only the author can catch. A transposed digit in a table. An author's name spelled wrong. An affiliation that has changed since submission. A figure panel labelled B where the caption says C. A reference that points at the wrong paper.
It is not an invitation to revise. Authors who are sent a proof and given no guidance will quite reasonably treat it as another draft, and will rewrite the discussion. Saying plainly what the proof is for, in the message that carries it, saves more time than any other single thing a publisher can do at this stage.
Send the real thing
A proof should be the typeset article, not a manuscript with tracked changes and not a list of queries against line numbers. Two reasons.
The first is that an author reading the typeset page sees what a reader will see, and catches things that are invisible in a manuscript: a figure that has lost its resolution, a table that has broken across a page in a way that makes a column unreadable, an equation that is set in the wrong variable.
The second is that a proof which looks nothing like the published article invites comments about the manuscript rather than the article, and most of those comments cannot be acted on.
Make the comments land somewhere
The common way to collect proof corrections is an email, which is the worst of the available options. It arrives as prose describing a location - "the second paragraph on page 4, the sentence beginning However" - and somebody has to translate each one back into a place in the article. On a long article with forty corrections, that translation is the work, and it is where corrections get missed.
A comment attached to the part of the article it refers to needs no translation. The editor sees the remark against the sentence, makes the change, and the next correction is already in front of them.
Give the link an expiry, not a password
An unpublished article is confidential, so a proof cannot simply be put at a guessable address. The usual answer is an account, which means the author has to register to read their own paper, and a third of them will not.
A private link that carries its own credential and stops working after a set period solves both halves: nobody needs an account, and the article is not left readable for ever at an address that was once emailed around.
A checklist for the editor, before the proof goes out
Most proof rounds that go badly go badly for reasons the editor could have caught first. Worth a minute each time:
- Every figure and table is cited in the text, and in order.
- Figure captions match the panels in the figure.
- Every reference in the list is cited, and every citation is in the list.
- Author names, affiliations and the corresponding author are as the author last gave them.
- The licence and copyright line are the journal's, and the right ones for this article.
- Received, revised and accepted dates are present and in order.
- The DOI on the first page is the one that will be registered.
Last reviewed
